A systematic review of clinicians' acceptance and use of clinical decision support systems over time

Abstract

Existing reviews have identified factors influencing Clinical Decision Support (CDS) adoption by clinicians in practice but overlook the dynamic and evolving nature of technology and users' needs over time. This review aimed to identify factors that influence early, mid-term, and sustained acceptance and use of CDS in hospital settings. Five databases were searched from 2007 to January 2024 and 67 papers were included. Factors were extracted and synthesised according to the time that data were collected following CDS implementation. Factors relating to the CDS intervention (e.g. utility) and inner setting (e.g. fit with workflows) were reported across all time periods. Perceived outcomes were more often identified in the first year of use, and individual factors after the first 6 months of use. Strategies to work around CDS limitations were reported 5 years after implementation. Our review provides guidance for developing, implementing, and supporting ongoing use of CDS systems.

Fig. 1

PRISMA flow diagram.

Fig. 2. Factors identified in CFIR domains over time.

m months, y years, n number of study time-points. The proportion of factors identified in each CFIR domain are presented relative to the total number of factors identified within each timeframe. Barriers, facilitators, and moderators were counted once per study time-point (see factor count calculation in Supplementary Table 3) and summed across all study time-points within each timeframe.

Fig. 3. Key constructs identified in CFIR domains over time.

Key constructs presented in this figure were identified as barriers, facilitators or moderators in over 25% of study time-points within a given timeframe. The proportion % and number () of study time-points where a construct was identified as a barrier (B), facilitator (F) or moderators (M) to CDS acceptance and use, relative to the total number of study timepoints identified within a given timeframe, are presented. For example, complexity appeared as a barrier in 4 study time-points conducted between 0–6 months, representing 14% of the total 28 study time-points included in this timeframe. Colour saturation was based on the proportion that constructs were reported within each timeframe (i.e. lighter = lower proportion, darker = higher proportion), with red gradients representing barriers, green representing facilitators and blue representing moderating factors. Constructs were counted once per study time-point (see construct count calculation in Supplementary Table 3) and summed across all study time-points within each timeframe.

Fig. 4. Barriers and facilitators identified in CFIR domains over time.

m months, y years, n number of study time-points. a The proportion of barriers identified in each CFIR domain are presented relative to the total number of barriers identified within each timeframe. Barriers were counted once per study time-point (see factor count calculation in Supplementary Table 3) and summed across all study timepoints within each timeframe. b The proportion of facilitators identified in each CFIR domain are presented relative to the total number of facilitators identified within each timeframe. Facilitators were counted once per study time-point (see factor count calculation in Supplementary Table 3) and summed across all study timepoints within each timeframe.